It’s very easy to use charts to support false arguments, distortions, omissions or outright lies. But you can use words and statistics too. If you want to deceive nothing will stop you. (Required reading: How To Lie With Charts and How to Lie with Statistics).
Simple lies are often easy to spot and not very interesting. More interesting are our biases. Germans call it weltanschauung (“world view”) and without it we wouldn’t be more than boring rational machines. Our biases help us to select the data and interpret it the way it makes sense to us, reinforcing our believes.
Lying with charts, if done properly (!), is more an act of omission (what you hide) than an act of commission (what you show). To better understand the differences, let me give you an example of how data visualization amateurs lie with charts.
In a post titled “charts can be deceiving”, E. D. Kain writes:
I’m not a huge fan of charts because I think they’re usually just used to create illusions and sales pitches.(…) Numbers don’t lie, but how we present them can make all the difference in the world.
Then he goes on and offers an example of how deceiving charts can be. I’ll recreate them for you. Chart A is the original chart, Chart B is his:
Jon writes about the flaws in both charts, so I’m not going to discuss them here.
It’s funny to see how a lack of action (the original chart accepts the Excel default scale) induces an over reaction (an absurd “theoretical” scale). Manipulating the y-axis scale is “How to Lie With Charts 101”.
Yes, charts can be deceiving. Words too. Numbers don’t lie? Bullsh*t. The political discourse is full of “illusions and sales pitches” and carefully selected and biased numbers. Yes, charts can be deceiving. It takes one to know one, I guess.
Deconstructing Lies with Charts
The original chart reveals a clear act of omission: how can you conclude anything relevant if you have no reference to compare the trend to?
So, let’s try to answer this question with the available data: are the wealthiest 1% of households getting a more favorable tax treatment, or not? First, we need some contextual data:
So, the wealthier you are the more taxes you pay; a downward trend is also visible across quintiles (although the highest quintile shows a slight increase over the last two years).
And what happens within the highest quintile?
Well, this is interesting: tax rates increased, but not for the top 1%. But the general rule is kept: the higher the income, the higher the tax rate.
What if we chart, not the tax rate but the change, assuming 1993=100? Again, some contextual data:
Tax rates for the lowest quintile declined sharply. And here is another general rule: the higher the tax rate, the less it changes.
Here is the detail chart for the highest quintile:
Well, it seems that at the very top some rules don’t apply, after all (surprise, surprise). The top 1% households did get a more favorable tax treatment after 1996, when compared to the top 5% and the top 10% households.
Takeaways
The world is never black and white, and your own shade of gray is as unique as your fingerprints. If you want to use charts to support your arguments, please don’t resort to scale tricks and make sure you add enough detail and contextual data.
There is no intrinsic objectivity in a chart, but if you want to support your story you should cover your bases and make sure it’s hard for someone else to come up with a different narrative. This is valid for charts but also for words and numbers.
Nice job. The last four charts are very interesting. In another analysis I’d seen the first and third, but the second and fourth show things I hadn’t seen before.
Excellent post. I’d like to point out that my own chart was not so much “over-reaction” as purposefully deceiving. I was merely attempting to draw contrast by comparison. Obviously the data presented in both charts was fundamentally flawed.
Really interesting comparison of top quintile by the way.
Dam it!
I was just about to launch in to a debate about how you needed to take the data in context of the whole population etc, etc, and there is was, what I was going to say anyway!
good post
Ross
nice conclusion. then again making sure viewers interpret the results you way you intend is more achieved by selecting the right variables, rather than following orthodox formatting rules.
Cool, the things you can “say” with a chart. However, it doesn’t matter how the state shows data its always twisted one way or other – to suit the message they want to pass. Not all charts are the same, for instance, Flynet are soon to launch a Dashboard Portal where small and medium enterprises can very simply choose their preferred KPI’s, enter their data, choose how they wish to see each KPI and are then presented with a fully functional dashboard. Charts in this case would be highly beneficial, offering a small business the ability to focu on those variables/ KPI’s most vital to their business.
Please have a look at http://www.hichert.com/de/success/check/201
HI-SUCCESS-Rule 4.4 Check scaling: No manipulation
4.4.1 Use proper visualization, do not cut axes